Correspondence
Glucocorticoids and Joint Destruction in Rheumatoid Arthritis
N Engl J Med 1995; 333:1569-1570December 7, 1995
- Article
To the Editor:
Kirwan et al. (July 20 issue)1 found that “in patients with early, active rheumatoid arthritis, prednisolone (7.5 mg daily) given for two years in addition to other treatments substantially reduced the rate of radiologically detected progression of disease.” Rheumatoid arthritis is a disease that lasts 20 to 25 years, and the ultimate goal of a therapeutic strategy is to reduce the cumulative pain, disability, side effects, economic impact, and mortality over the course of the illness. Long-term, prospective observational studies of prednisone use show very disquieting results. With the use of average prednisone doses (6.9 mg per day) similar to those given by Kirwan et al. and after adjustment for appropriate covariates, prednisone therapy is associated with markedly increased disability,2 a doubled risk of hospitalization for gastropathy due to nonsteroidal antiinflammatory drugs,3 and a twofold increase in standardized mortality rates.4 Quantitative estimation of side-effect (toxicity index) scores shows higher scores for prednisone than for nearly all other agents used to treat rheumatoid arthritis.5
We hope that the study by Kirwan et al. will not lead to a major increase in the use of low-dose prednisone in the management of rheumatoid arthritis in the unsupported belief that long-term outcomes will be improved. Win a battle, lose a war.
5 ReferencesJames F. Fries, M.D.
Gurkirpal Singh, M.D.
Stanford University Medical Center, Palo Alto, CA 943041
Kirwan JR, Arthritis and Rheumatism Council Low-Dose Glucocorticoid Study Group
Full Text | Web of Science | Medline2
Sherrer YS ,Bloch DA ,Mitchell DM ,Roth SH ,Wolfe F ,Fries JF . Disability in rheumatoid arthritis: comparison of prognostic factors across three populations. J Rheumatol 1987;14:705-709
Web of Science | Medline3
Fries JF ,Williams CA ,Bloch DA ,Michel BA . Nonsteroidal anti-inflammatory drug-associated gastropathy: incidence and risk factor models. Am J Med 1991;91:213-222
CrossRef | Web of Science | Medline4
Wolfe F ,Mitchell DM ,Sibley JT , et al. The mortality of rheumatoid arthritis. Arthritis Rheum 1994;37:481-494
CrossRef | Web of Science | Medline5
Fries JF ,Williams CA ,Ramsey DR ,Bloch DA . The relative toxicity of disease-modifying antirheumatic drugs. Arthritis Rheum 1993;36:297-306
CrossRef | Web of Science | Medline
To the Editor:
I am concerned that the conclusions of Kirwan et al. may not be valid. The placebo group included, by chance, all the patients with the most severe disease. Dealing with highly skewed data and differences in the severity of disease at base line requires careful consideration of the statistical analyses applied.
The authors do not explain how they performed a log transformation of data that included patients with a Larsen score of zero. With so many patients starting with a Larsen score of zero, log transformation could never compensate for the skewed distribution and result in a near-normal distribution.
Analyzing the change in the log-transformed Larsen score is equivalent to analyzing the ratio of Larsen scores at two years and at the outset of the study. This is not a valid measure of disease progression. Was an increase in the Larsen score from 1 to 3 (ratio, 3) regarded as a worse outcome than a change from 10 to 25 (ratio, 2.5)? Furthermore, the mean (±SD) change in the log-transformed Larsen score is given as 0.02±0.43 for the prednisolone group. If they were derived from a near-normal population, these figures would indicate that in the prednisolone group, as many patients had radiologic improvement as had deterioration. On the basis of clinical experience, the parametric analyses performed seem inappropriate. The appropriate analysis for these data would use an analysis of covariance, with regression of the final score according to both treatment group and base-line score.
I am also concerned about the analysis of new erosive change in 212 hands. Observations on the two hands of the same patient are not independent and cannot be treated as such. The analysis should be repeated using the number of patients in whom erosive changes occurred.
Duncan Porter, B.M., B.Ch.
Gartnavel General Hospital, Glasgow G12 0YN, United KingdomTo the Editor:
Kirwan et al. do not look at bone-mass measurements, nor do they mention the possible effects of prednisolone on bone density. Prednisolone in low doses similar to the ones used by Kirwan et al. affects bone mass adversely in patients with rheumatoid arthritis.1,2 We have used densitometry of the hand bones to monitor the progression of rheumatoid arthritis.3 In a prospective longitudinal study of 42 patients, we have shown that hand-bone mass decreases and Larsen scores based on x-ray films of the hands worsen in patients with early rheumatoid arthritis, despite significant improvement in clinical status over a one-year period.4 There was no correlation between the worsening of the Larsen scores and the decreasing hand-bone mass. The slower rate of deterioration in Larsen scores in the study by Kirwan et al. may not mean a slower rate of deterioration in hand-bone mass; in fact, it may mean quite the reverse.
Studies examining major determinants of bone mass in patients with rheumatoid arthritis have identified disease activity (as measured by an elevated acute-phase response) and use of oral corticosteroids as the most important detrimental factors.1 Throughout the study by Kirwan et al., patients in the prednisolone group had disease activity (as measured by the acute-phase response) comparable to that in the control group. It is therefore highly likely that over a two-year period, patients in the prednisolone group may have lost more bone mass than those in the placebo group.
The harmful effects of corticosteroids on bones should be carefully weighed against their possible benefits when one is prescribing corticosteroids to patients with rheumatoid arthritis.
4 ReferencesA.A. Deodhar, M.D.
Oregon Health Sciences University, Portland, OR 97201A.D. Woolf, B.Sc.
Royal Cornwall Hospital, Truro TR1 2HZ, United Kingdom1
Laan RF ,van Riel PL ,van de Putte LB . Bone mass in patients with rheumatoid arthritis. Ann Rheum Dis 1992;51:826-832
CrossRef | Web of Science | Medline2
Laan RF ,van Riel PL ,van de Putte LB ,van Erning LJ ,van't Hof MA ,Lemmens JA . Low-dose prednisone induces rapid reversible axial bone loss in patients with rheumatoid arthritis: a randomized controlled study. Ann Intern Med 1993;119:963-968
Web of Science | Medline3
Deodhar AA ,Brabyn J ,Jones PW ,Davis MJ ,Woolf AD . Measurement of hand bone mineral content by dual energy x-ray absorptiometry (DXA): development of the method, and its application in normal volunteers and in patients with rheumatoid arthritis. Ann Rheum Dis 1994;53:685-690
CrossRef | Web of Science | Medline4
Deodhar AA, Brabyn J, Jones PW, Davis MJ, Woolf AD. Longitudinal study of hand bone densitometry in rheumatoid arthritis. Arthritis Rheum (in press).
Author/Editor Response
Dr. Kirwan replies:
To the Editor: The statistical analysis of the data in our study was undertaken to test specific hypotheses defined in advance within the study protocol. Several approaches were taken in relation to the skewed distribution of the Larsen scores for the severity of disease on x-ray films, including parametric analysis of two types of log transformation (of the raw scores and of the change in scores during treatment) and nonparametric analysis of the raw data. Although the absolute values of the differences between groups and their associated P values differed slightly for each analysis, all results pointed to a significant treatment effect with a P value of less than 0.05 in all cases. We repeated the analyses omitting the two highest-scoring patients, and the results were equally significant. The analysis of hands with erosions was simply a comparison of proportions. The hands of a patient were considered separately because we have no information about the state of the patient's other joints, and it would therefore be impossible to classify a patient as having erosive disease or nonerosive disease.
The secondary hypotheses of the study related to clinical outcome, which was illustrated in Figure 2 of our article. Multiple comparisons made in this analysis were not corrected by the Bonferroni method, and some P values would be less than 0.05 if adjusted in this way, but taken together the results show an overall short-term clinical benefit measured in a variety of ways in the prednisolone group. Furthermore, existing knowledge that glucocorticoids do not make symptoms worse argues that the significance tests could have been one-tailed. Such an approach would substantially reduce the P values.
Long-term use of steroids is associated with increased morbidity in patients with rheumatoid arthritis, as suggested by Fries and Singh. Unfortunately, there is no clear evidence that this is a causal relation rather than the association of a treatment used in severe disease with the more severe outcomes associated with that degree of disease severity. Controlled clinical trials such as ours are able to distinguish between cause and simple association. On the basis of our evidence, treatment can be recommended for only two years. Further controlled studies will be required to justify longer-term treatment.
The results of measurements of bone mineral density in a subgroup of the patients in our study are currently being analyzed. The power of this analysis will be low, and there is a need to measure bone mineral density in more patients in randomized, controlled trials.
John R. Kirwan, M.D.
Bristol Royal Infirmary, Bristol BS2 8HW, United Kingdom- Citing Articles (3)
Citing Articles
1
John R. Kirwan. (2001) SYSTEMIC LOW-DOSE GLUCOCORTICOID TREATMENT IN RHEUMATOID ARTHRITIS. Rheumatic Disease Clinics of North America 27:2, 389-403
CrossRef2
Ernest HS Choy. (1998) New prospects for the treatment of rheumatoid arthritis. Expert Opinion on Investigational Drugs 7:7, 1087-1097
CrossRef3
W. Grassi, R. Angelis, C. Cervini. (1998) Corticosteroid prescribing in rheumatoid arthritis and psoriatic arthritis. Clinical Rheumatology 17:3, 223-226
CrossRef
